Abstract
Background subtraction is a method typically used to segment moving vehicles in image sequences taken from a static camera by comparing each new frame with a model of the background scene. This paper presents a robust background subtraction algorithm which reduces the influence of illumination changes and shadows and adapts to rapid changes in the traffic scene. A statistical background modeling method is presented, which is based on a histogram at the pixel level and produces a color model from a series of frames. For foreground detection, we propose the Choquet integral to fuse the three color-component similarity measures and a texture similarity measure based on a uniform local binary pattern. Finally, we propose a new adaptive background maintenance method. The experimental results for several dataset videos show that the proposed method is more efficient, robust, and accurate than classical approaches.
